PUTTING NUMBERS TO GEOLOGY
The Second Glossop Lecture
Putting Numbers to Geology – an Engineer’s Viewpoint
Evert Hoek
3034 Edgemont Boulevard, North Vancouver, British Columbia, Canada V7R 4X1
Abstract
Assigning numbers to geology requires a delicate
balance between the commonly held opinion that
geology cannot be quantified and the over-
optimistic view that every physical quantity can
be described in precise mathematical terms. In
reality, many geological characteristics cannot be
quantified precisely and intelligent guesses based
upon experience and logical arguments are the
best that can be hoped for.
This paper explores the processes used to
make some of these guesses and describes how
the results are then applied to engineering design.
It is shown that, with care, rational engineering
decisions can be made in spite of the limitations
of the input data. In recent years the development
of computer hardware and software has made it
much easier to investigate the influence of ranges
of values for each of the input parameters. How-
ever, care has to be taken that the design is driven
by sound geological reasoning and rigorous engi-
neering logic rather than by the very attractive
images that appear on the computer screen.
Introduction
Professor Peter Fookes, in the First Glossop lec-
ture (Fookes 1997), gave an excellent description
of the numerous steps required in the develop-
ment of a Geological Model. This model, whether
conceptual, hand-drawn or in the form of a com-
puter generated three-dimensional solid model, is
the basic building block upon which the design of
any major construction project must be based. A
good geological model will enable the geologists
and engineers involved in the project to under-
stand the interactions of the many components
that make up the earth’s crust and to make ra-
tional engineering decisions based on this under-
1
standing. On projects where an adequate geological
model does not exist, decisions can only be made on an
ad hoc basis and the risks of construction problems due
to unforeseen geological conditions are very high.
In this, the Second Glossop lecture, I would like to
take the process of design to the next step. I will attempt
to describe how an engineer puts numbers to the largely
qualitative model described by Fookes. Many geologists
are uncomfortable with this requirement to assign num-
bers to geology and many will contend that geological
materials, not being man-made like steel or concrete,
cannot be quantified. While I have some sympathy with
these views, I have to face the reality that engineering
design requires numbers in the form of in situ stress,
pore water pressure, rock mass strength and deformation
modulus. These numbers are required for the calculation
of the stability of slopes, the bearing capacity of foun-
dations, the support capacity for underground excava-
tions and the movement of groundwater contaminants.
Without these numbers the process of engineering de-
sign is not possible.
Of course rock and soil are not man-made and their
properties can vary greatly over short distances. The
interactions of different components in a rock mass can
be very complex and these interactions are difficult to
quantify. These variations must be recognised and in-
corporated into the numbers themselves and the use to
which the numbers are put in the engineering design
process. Quoting a rock mass classification value to
three decimal places betrays a complete lack of under-
standing of the process of quantifying rock mass prop-
erties. On the other hand, assigning excessively large
ranges to each parameter can result in equally meaning-
less results.
A good engineering geologist and a good geotechni-
cal engineer, working as a team, can usually make real-
istic educated guesses for each of the parameters re-
quired for a particular engineering analysis. It is the
selection of reasonable values for the parameters and the
choice of appropriate engineering design methods that I
wish to explore in this paper.
2 E. HOEK
Help for artistically challenged geologists
The three-dimensional block drawings and sec-
tions included in the written version of the first
Glossop Lecture, prepared by or with the assis-
tance of Mr G. Pettifer, are miniature master-
pieces of geological art. If only such drawings
were available on all construction sites.
Unfortunately, I have to say that in my thirty-
five odd years of consulting around the world I
have seldom come across geological drawings
that come close to the these in terms of clarity of
presentation and transmission of useful engineer-
ing geology information. The converse is gener-
ally the case and I have spent many uncomfort-
able hours attempting to decipher geological plans
and sections of less than adequate quality. Of
course, it is not the artistic ability of the geologist
that determines that accuracy of the geological
interpretations being presented but it certainly
helps when the drawings are well executed,
clearly captioned and approximately to scale.
Help for artistically challenged geologists is
on the way in the form of computer generated
three-dimensional solid models. Such models are
now relatively common in mechanical and struc-
tural engineering and even in the medical field.
The models of greatest interest to geologists were
developed to meet the needs of the mineral explo-
ration geologists in their efforts to define the three
dimensional shapes and ore grade distribution of
sub-surface mineral deposits. For many years
these geologists have used sophisticated statistical
techniques and trend surface analysis to interpo-
late and extrapolate between borehole intersec-
tions. The evolution into three-dimensional com-
puter modelling was a natural step.
The mining industry has embraced these com-
puter modelling techniques and such models can
now be found in mine planning and geotechnical
departments as well as in the offices of the explo-
ration and mining geologists.
One of the most spectacular examples of such
a model has been constructed by the Geotechnical
Group of the Chuquicamata open pit copper mine
in northern Chile, illustrated in Figure 1. An ex-
ample of a typical three-dimensional block model
is illustrated in Figure 2. The 1998 shell of the
Chuquicamata mine, showing the geological units
exposed in the walls, is illustrated in Figure 3. In
this case the computer operators are the geologists
themselves and it is not unusual to see a geologist come
in from the field and sit down immediately to enter the
latest data into the model. This ensures that the model
reflects the understanding and interpretation of the ge-
ologists and that it is not simply an illustration prepared
by a computer technician who may not understand the
on-going thinking that goes into building the geological
model.
The advantages of these three-dimensional computer
generated models are enormous. The model can be ro-
tated and viewed from any direction, enlarged, sectioned
and components can be removed or added at will. Trend
surfaces representing interpolations or extrapolations
between boreholes can be adjusted to fit the geologist’s
understanding of the tectonic processes involved in the
formation of the rock mass. Work is now going on to
take data from one of these models and to feed it di-
rectly into limit equilibrium slope stability analyses or
numerical analyses of the stress and failure conditions
around underground excavations.
The current cost of the hardware and software re-
quired for the generation of these three-dimensional
models is approximately £50,000. This places it outside
the range of all but the very largest civil engineering
projects. However, with dramatic advances in computer
software and the ever decreasing cost of computer
hardware, it is conceivable that installations costing one
tenth of the current system costs will be available within
a few years. This would put these systems within reach
of most agencies or consulting organisations with the
need to interpret and present engineering geology data. I
look forward with eager anticipation to the day when I
see one of these models being used on a civil engineer-
ing project.
The geotechnical engineering design process
The end product of the work carried out by a geotechni-
cal engineer is generally the complete design of a slope,
a foundation or an underground excavation. An example
of a typical flow path for a geotechnical engineering
design, adapted from Hoek and Brown (1980), is illus-
trated in Figure 4. In this case, the design is for an un-
derground excavation but a similar diagram can be con-
structed for any other structure for which the geotechni-
cal engineer is responsible
From this figure it will be obvious that the design
process progresses from a largely qualitative prelimi-
nary assessment of potential problems to a highly quan-
titative analysis of support capacity and excavation per-
formance for the situations that require such an analysis.
 PUTTING NUMBERS TO GEOLOGY
Fig 1: Arial view of the Chuquicamata open pit copper mine in northern Chile.

3
4 E. HOEK

Fig. 2: Example of a computer generated three-dimensional solid model of the rock mass in which the
Chuquicamata open pit copper mine in northern Chile is being mined.

Fig. 3: Chuquicamata open pit mine in 1998 showing the geological units exposed in the walls of the 750 m
deep pit. Figures 2 and 3 were prepared by Mr Ricardo Torres of the Chuquicamata Geotechnical Group
using the program Vulcan1.

1
Available from Maptek Perth, 92 Roe Street, Northbridge, Western Australia 6003, Phone: + 61 8 9328
4111, Fax: + 61 8 9328 4422, email: info@perth.maptek.com.au
 PUTTING NUMBERS TO GEOLOGY 5

Fig. 4: Flow path for the geotechnical design of underground excavations in rock. (Hoek and Brown 1980).
6 E. HOEK
Note that the engineering design process need
only be taken as far as necessary to satisfy the
designer that the requirements of safety and sta-
bility have been met. It may be possible, on the
basis of a very simple semi-quantitative analysis,
to conclude that there are no conditions likely to
lead to instability and to terminate the design at
this point. On the other hand, in cases where the
structural conditions are very unfavourable or
where the rock mass strength is very low com-
pared to the in situ stresses, a very detailed nu-
merical analysis may be required.
In complex cases it may be necessary to run
the numerical analysis concurrently with con-
struction and adjust the excavation sequence and
support systems to satisfy the design requirements
established by back-analysis of the observed ex-
cavation behaviour.
Note that the geological model is a dynamic
tool that changes as more information is exposed
during the excavation process. It is only for very
simple geological environments that the geologi-
cal model can be established early in the site in-
vestigation and design process and left unaltered
for the remainder of the project. The more usual
condition is that the model is continually refined
as the project progresses through the various
stages of design and construction.
Preliminary project feasibility assessment
During the very early stages of project evaluation
and design, when practically no quantitative in-
formation is available and when the geological
model is fairly crude, the design process relies
heavily on precedent experience and very general
rules of thumb. For example, in evaluating three
alternative highway routes through mountainous
terrain, the engineering geologist or geotechnical
engineer would look for routes with the minimum
number of unstable landforms, ancient landslides,
difficult river crossings and the minimum number
of tunnels. Simple common sense says that all of
these factors represent problems and the potential
for increased cost.
This may sound a trivial example but it is
amazing how often a highway will be laid out by
transportation engineers with more concern for
lines of sight and radii of curves than for the
geological conditions which happen to occur
along the route. It is then up to the engineering
geologists and geotechnical engineers to sort out the
problems and, where necessary, to propose an alignment
that is more appropriate for the geological conditions.
Precedent experience is also an important considera-
tion at this stage of the design process. When evaluating
the potential problems along a proposed tunnel route it
is very useful to visit and to talk to engineers and con-
tractors who have worked on tunnels in similar geologi-
cal conditions within a few tens of kilometres of the site,
if such tunnels exist.
Care has to be exercised in how this precedent expe-
rience is interpreted and applied. I remember visiting an
open pit mine in the United Kingdom many years ago
and asking why the slopes had been designed at the un-
usual angle of 53 degrees. The answer I received was
that the company’s mines in the United States seemed to
operate successfully at this angle – hardly an appropriate
extrapolation by any stretch of the imagination.
During the preliminary design stage, the engineer is
probably less important than the geologist. The engineer
is there to convey the general requirements and con-
straints of the project and it is up to the geologist, based
on the geological model, to provide the qualitative as-
sessment of whether these conditions can easily be met
or whether it would be better to look for another site.
Preliminary engineering evaluation
Once the qualitative process described above has been
exhausted and the options have been narrowed down to
one or two, it may become necessary to move into a
more quantitative process in which the engineer starts to
assume the leading role in the design process. It is at this
stage in the design process (and, in my opinion, only at
this stage) that classification schemes play an important
role.
These classifications, based upon experience and the
back analysis of a large number of case histories, at-
tempt to quantify the general rock mass conditions in
terms of relatively simple numerical ratings. The final
‘score’ is then used to provide guidance on tunnel sup-
port, slope stability, the problems of excavating rock
masses or the ease with which a rock mass will cave in a
block caving mining operation. The rock mass classifi-
cation systems commonly used in the English language
world have been summarised by Bieniawski (1989) and
it is not my intention to discuss these classifications
further here. Incidentally, there are at least seven differ-
ent rock mass classification systems in use in Japan and
probably similar numbers in other non-English speaking
countries.
Table 1: Rockfall Hazard Rating System. After Pierson and van Vickle (1993).
8 E. HOEK

Table 2: Example of the application of the Rockfall Hazard Rating System

Category Description Points

Slope height 30 m 81
Ditch effectiveness Limited catchment 27
Average vehicle risk 50% of the time 9
Percentage of decision sight distance Very limited sight distance, 40% of low design value 81
Roadway width, including paved 28 feet / 8.5 m 27
shoulders
Geologic character – Case 1 Discontinuous joints, adverse orientation, Planar 27
Block size / quantity of rockfall 3 ft (1.3 m) / 12 cu. yards or cu. metres 81
Climate and presence of water on Moderate precipitation or short freezing periods or 9
slope intermittent water on slope
Rockfall history Many falls 27
Total score 369

A classification system that is probably almost
completely unknown in the United Kingdom but
which, for me, embodies the essential elements of
a good classification system for preliminary engi-
neering design is the ‘Rockfall Hazard Rating
System’. This system was developed by the Fed-
eral Highway Administration in the United States
for the preliminary evaluation of rockfall hazards
and the allocation of priorities for remedial work
(Pierson and van Vickle 1993). The key elements
of this rating system are contained in the table
reproduced as Table 1. Detailed instructions and
examples on the evaluation of each of the nine
components of the system are given in the FHWA
manual.
I like this classification because it is based on
a set of simple visual observations, most of which
can be carried out from a slow moving vehicle as
would be required for the preliminary evaluation
of miles of mountain highway. The system also
contains all the components required for a com-
plete engineering evaluation of the risks to the
public. These include highway design factors as
well as geometrical and geotechnical factors, all
presented in clear and unambiguous terms.
An example of a typical rockfall hazard
evaluation, based on this system, is given in Table
2. The authors of the FHWA manual give no di-
rect instructions on how the total score obtained from
this rating system should be used. It is intended for use
as a tool to assist management in the allocation of re-
sources and these decisions will vary from state to state.
From personal discussions with one of the authors I
learned that, in the State of Oregon, slopes with a rating
of less than 300 are assigned a very low priority while
slopes with a rating of more than 500 are identified for
urgent remedial action.
Returning to the question of the preliminary evalua-
tion of a construction project, the aim should be to di-
vide the problems into a series of approximate catego-
ries, depending upon the severity of each problem.
Whatever numerical process is used, these categories
should be treated as approximate guidelines rather than
absolute design values. The whole purpose of the pre-
liminary evaluation is to decide which components jus-
tify additional site investigations and analysis. The de-
tailed design follows later.
Detailed engineering design
Having identified those components of a construction
project that require detailed analysis, the next step is to
select the appropriate method of analysis and the input
data required for this analysis. There are too many geo-
technical problems and methods of analysis for me to
 PUTTING NUMBERS TO GEOLOGY
cover in this paper so I will deal with only one -
the design of underground excavations in weak
rocks.
In the context of this discussion I will define
rock as weak when the in situ uniaxial compres-
sive strength is less than the in situ stress level.
Hence, a jointed rock mass with a uniaxial com-
pressive strength of 3 MPa will behave as a weak
rock at depths of more than about 120 m. Under
these conditions a tunnel would begin to show the
first signs of stress induced failure.
In order to carry out a meaningful analysis of
the stresses induced by the excavation of a tunnel
or cavern it is necessary to estimate the in situ
stresses in the rock mass and also the properties of
the rock mass.
Estimates of in situ stress
Of all of the quantities that the geotechnical engi-
neer is required to estimate or to measure, the in
situ stress field in a rock mass is one of the most
difficult. The vertical stress can be approximated,
to an acceptable level of accuracy, by the product
of the depth below surface and the unit weight of
the rock mass. On the other hand, the horizontal
stresses of interest to civil engineers are influ-
enced by global factors such as plate tectonics and
also by local topographic features.
Zoback (199) described the World Stress Map
project that was designed to create a global data-
base of contemporary tectonic stress data. The
data included in this map were derived mainly
from geological observations on earthquake focal
mechanisms, volcanic alignments and fault slip
interpretations.
The results included in this map are very in-
teresting to geologists but are of limited value to
engineers concerned with the upper few hundred
metres of the earth’s crust. The local variations in
the in situ stress field are simply too small to
show up on the global scale.
A more useful basis for estimating horizontal
in situ stresses was proposed by Sheorey (1994).
He developed an elasto-static thermal stress
model of the earth. This model considers curva-
ture of the crust and variations of elastic con-
stants, density and thermal expansion coefficients
through the crust and mantle. A plot of the ratio of
horizontal to vertical stress predicted by Sheo-
rey’s analysis, for a range of horizontal rock mass
9
deformation moduli, is given in Figure 5. This plot is
very similar in appearance to that derived by Hoek and
Brown (1980) on the basis of measured in situ stresses
around the world. While this similarity does not consti-
tute a proof of the correctness of Sheorey’s solution, it is
at least comforting to find this correlation between the-
ory and observations.
k = ra tio of h orizontal to v e rtic al in s itu s tress
0 1 2 3 4
0
500
D e pth b elo w surfa ce z - m etres
1000 Horizo ntal defo rm ation
modu lus Eh (G Pa)
10
1500
25
50
2000
75
100
2500
Sh eo rey's eq ua tio n
k = 0 .25 + 7 E h (0.0 01 + ( 1/z))
3000
Fig. 5. Ratio of horizontal to vertical in situ stress versus
depth below surface. (Sheorey 1994)
Note that neither Sheorey’s equation nor the trends
established by Hoek and Brown account for local topo-
graphic influences on the in situ stress field. Hence,
when making estimates of the in situ stress field in a
mountainous area, adjustments must be made to account
for these topographic factors. For example, the general
relationships discussed above may indicate a horizontal
stress of approximately twice the vertical stress for the
rock mass at a depth of 300 m. In deciding upon the in
situ stresses to be applied to the analysis of an under-
ground powerhouse to be located at this depth in the
side of a steep valley, the horizontal stress at right an-
gles to the valley axis could be reduced to a value equal
to the vertical stress. This would account for the stress
relief due to the down-cutting of the valley. No such
stress relief would occur parallel to the valley axis and
so the horizontal stress in this direction would be kept at
twice the vertical stress.
10 E. HOEK

In carrying out an analysis of the stresses in- Estimates of rock mass properties
duced by the creation of an underground excava-
tion, it is prudent to consider a range of possible Hoek and Brown (1980) proposed a methodology for
in situ stresses. In the example discussed above, estimating the strength of jointed rock masses. This
the horizontal stress at right angles to the valley technique has been refined and expanded over the years
axis could be varied from one half the vertical and the latest version is described in a recent paper and
stress to twice the vertical stress. The stress par- technical note. (Hoek and Brown 1997, Hoek 1998).
allel to the valley could be varied from a mini- The basic input consists of estimates or measure-
mum value equal to the vertical stress to a maxi- ments of the uniaxial compressive strength ( σ ci ) and a
mum value of three times the vertical stress. An material constant (mi) that is related to the frictional
exploration of the effects of all possible combina- properties of the rock. Ideally, these basic properties
tions of these stress values would give a good should be determined by laboratory testing as described
indication of whether or not these in situ stresses by Hoek and Brown (1997) but, in may cases, the in-
would be critical to the design of the underground formation is required before laboratory tests have been
excavations. In cases where a preliminary analysis completed. To meet this need, tables that can be used to
indicates that the design is very sensitive to the in estimate values for these parameters are reproduced in
situ stresses, measurement of the in situ stresses Tables 3 and 4.
has to be considered a priority in the ongoing site
investigation and design process.
Table 3: Field estimates of uniaxial compressive strength.
Uniaxial Point
Comp. Load Field estimate of strength
Grade* Term Strength Index Examples
(MPa) (MPa)
R6 Extremely > 250 >10 Specimen can only be chipped Fresh basalt, chert, diabase,
Strong with a geological hammer gneiss, granite, quartzite

R5 Very 100 - 250 4 - 10 Specimen requires many blows of Amphibolite, sandstone, basalt,
strong a geological hammer to fracture it gabbro, gneiss, granodiorite,
limestone, marble, rhyolite, tuff

R4 Strong 50 - 100 2-4 Specimen requires more than one Limestone, marble, phyllite,
blow of a geological hammer to sandstone, schist, shale
fracture it

R3 Medium 25 - 50 1-2 Cannot be scraped or peeled with a Claystone, coal, concrete, schist,
strong pocket knife, specimen can be shale, siltstone
fractured with a single blow from a
geological hammer

R2 Weak 5 - 25 ** Can be peeled with a pocket knife Chalk, rocksalt, potash

with difficulty, shallow indentation
made by firm blow with point of a
geological hammer

R1 Very 1-5 ** Crumbles under firm blows with Highly weathered or altered rock
weak point of a geological hammer, can
be peeled by a pocket knife

R0 Extremely 0.25 - 1 ** Indented by thumbnail Stiff fault gouge

Weak

*Grade according to Brown (1981).

**Point load tests will give highly ambiguous results on rocks with a uniaxial compressive strength of less than 25 MPa.
 PUTTING NUMBERS TO GEOLOGY 11

Table 4: Values for the constant mi for intact rock, by rock group. Note that the
values in parenthesis are estimates.

The most important component of the Hoek-
Brown system is the process of reducing the ma-
terial constants σ ci and mi from their ‘laboratory’
values to appropriate in situ values. This is ac-
complished through the Geological Strength In-
dex GSI that is defined in Figure 6.
In the context of this paper, the GSI is a real case of
putting numbers to geology. It has been developed over
many years of discussions with engineering geologists
with whom I have worked around the world. Careful
consideration has been given to the precise wording in
each box and to the relative weights assigned to each
combination of structural and surface conditions.
12 E. HOEK

Fig. 6: Geological Strength Index GSI on the basis of geological observations.

 PUTTING NUMBERS TO GEOLOGY 13

The version of the GSI chart presented in Fig- m i

55 35
ure 6 contains two new rows that have not yet 30
been published elsewhere. The top row on ‘intact 50 25
20
or massive’ rock is the result of work in Chile on 45 16
cemented breccias that behave very much like 12
weak concrete (personal communication from Dr 40 10

F rictio n a n g le - d e g re e s
Antonio Karzulovic). The bottom row on ‘foli- 35 7
ated/laminated/sheared’ rock has been inserted to
deal with very poor quality phyllites encountered 30 5

in Venezuela (personal communications from 25

Professors Rudolpho Sancio and Daniel Salcedo)
20
and the weak schists being tunnelled through for
the Athens Metro (Hoek, Marinos and Benissi 15
1998). It is probable that this figure will continue
10
to evolve as experience is gained in the use of
GSI for estimating rock mass properties in the 5
wide range of geological environments to which it 0
is being applied. 0 10 20 30 40 50 60 70 80 90 1 00
Based on intuition, experience and the back
G e olo g ical S tren g th In d e x G S I
analysis of a number of case histories, relation-
ships have been developed between GSI, σ ci and Fig. 8: Friction angle versus GSI.
mi and the various rock mass properties required 180 1 00
for engineering analyses,. These relationships, σ ci
E m (GPa ) = ⋅ 10 ((GSI −10) 40)
160 100

U n ia xia l co m p re ssive s tre n g th

described in detail by Hoek and Brown (1997), 75
D e fo rm a tio n m o d u lu s E m - G P a

have been used to generate the charts for cohe- 140

o f in ta ct ro ck σci - M P a
sion, friction angle and modulus of deformation 50
given in Figures 7, 8 and 9. 120

These charts can be used to obtain approxi- 100

mate values for in situ properties. It is an absolute 25

requirement that the engineer making these esti- 80

mates should check their appropriateness by back 60 10

analysis of the measured or observed excavation
behaviour, once construction commences. 40 5

0. 40 20 1
C o he s iv e s tren g th / U n ia x ia l s tre n g th o f in ta c t ro c k

0. 30
0
0. 20
0 10 20 30 40 50 60 70 80 90 100
G eo lo g ica l S tre ng th In de x G S I
0. 10
Fig 9: Deformation modulus versus GSI.
0. 05
0. 04 Practical example
0. 03
m A 27 km long, 10 m internal diameter concrete-lined
i
0. 02
35
30 headrace tunnel is currently under construction as part
25
20
0. 01
of the 1500 MW Nathpa Jhakri hydroelectric project on
16
12 the Satluj river in Himachel Pradesh, India. The rock
10
7 masses through which the tunnel passes are either
5 0 .0 05
metamorphic, consisting of gneisses, schists, quartzites
0 10 20 30 40 50 60 70 80 90 1 00
and amphibolites or igneous consisting of granites and
G e olo gic al S t re ng th In d ex G S I
pegmatites. The engineering geological conditions asso-
Fig. 7: Cohesive strength versus GSI.
14 E. HOEK
ciated with the project have been evaluated by the
Geological Survey of India (Geological Survey of
India 1988, Jalote et al 1996) on the basis of sur-
face mapping, exploration boreholes and a few
exploration adits. Excellent maps and sections
were available before the commencement of un-
derground excavation. In addition to conventional
descriptive and structural maps, the rock mass has
been classified in terms of Bieniawski’s RMR
system (Bieniawski 1989), Barton, Lien and
Lunde’s Q system (Barton et al 1974) and the GSI
system described above.
At the time of writing (May 1998), the bulk of
the tunnel excavation has been completed and the
prediction of tunnelling conditions provided by
the Geological Survey of India has proved to be
accurate and a useful guide to the steps to be
taken in excavation and support. One of the sec-
tions still to be completed is a 360 m long stretch
through the Daj Khad shear zone. It is this part of
the tunnel that I wish to discuss. The dramatic
impact of the Daj Khad shear zone on the stability
of the tunnel top heading is illustrated in Figure
10. This shows a closure in excess of one metre
Fig. 10: Large convergence in the Nathpa Jhakri headrace tunnel top heading due to the influence of the Daj
Khad shear zone.
due to the heavy loads being imposed on the support
system.
The rock mass in the vicinity of the Daj Khad shear
zone is predominantly quartz mica schist with some
serecite schist and a few gneiss bands and one amphi-
bolite zone. The shear zone itself comprises a number of
steeply dipping seams of fractured blocky rock with
kaolinised and sericitised material. The uniaxial com-
pressive strength of the schist that makes up the bulk of
the rock mass is approximately 10 MPa under the satu-
rated conditions that occur at the tunnel depth of be-
tween 200 and 300 m through this zone. The value of
the rock mass constant mi has been assumed equal to 10
for the entire zone (see Table 4). The variation of the
Geological Strength Index GSI through the rock mass
associated with the Daj Khad can be represented by a
truncated normal distribution defined by a mean value
of 27, a standard deviation of 7, a minimum value of 6
and a maximum value of 45. This distribution is based
on studies carried out by Geodata S.p.A. of Turin, con-
sultants to the Nathpa Jhakri Joint Venture, the con-
tractors on this stretch of headrace tunnel. The method-
ology employed by Geodata in arriving at this distribu-
tion has been described in a recent paper by Russo et al
(1998).
 PUTTING NUMBERS TO GEOLOGY 15

Based upon this GSI distribution and assuming 1. 0

that the vertical in situ stress is uniformly distrib- 0. 9

uted with a minimum of 5.4 MPa and a maximum
of 8.1 MPa, corresponding to depths below sur- 0. 8
face of 200 and 300 m, a Monte Carlo simulation 0. 7
has been carried out to determine the extent of the

T unn el con ve rge nce

T un nel diam e te r
plastic zone and the convergence of the rock mass 0. 6
surrounding the 10 m diameter tunnel. This cal- 0. 5
culation is too detailed for inclusion in this publi-
cation but the equations used to set up the spread- 0. 4
sheet for the simulation are described in Hoek and 0. 3
Brown (1997) and Hoek (1998). The results of the
simulation are plotted, in dimensionless form, in 0. 2
Figures 11 and 12. Note that these plots are for an
0. 1
unsupported tunnel.
It is evident, from the plots given in these fig- 0. 0
ures, that the size of the plastic zone and the con- 0. 0 0. 1 0. 2 0. 3
vergence of the tunnel both show dramatic in- Uniaxial c om p ress ive stren gth of r ock mas s
creases when the uniaxial compressive strength of In situ stres s
the rock mass falls below about one tenth of the in
situ stress. Unless adequate support is provided, Fig. 12: Tunnel convergence versus ratio of uniaxial
the tunnel will almost certainly collapse for the compressive strength of the rock mass to in situ stress.
lowest quality rock conditions under the highest
in situ stresses. These findings are consistent with In passing, it is worth mentioning that trends such as
the results of as yet unpublished research on tun- this are of great value to geotechnical engineers. If a
nelling in weak rocks. I have found that the very trend is found to be consistent over a wide range of con-
unstable conditions develop in unsupported tun- ditions, this usually indicates that some basic law is at
nels of almost any shape for rock mass strengths work and, if this law can be isolated, it may be possible
less than 0.1 to 0.2 of the maximum in situ stress. to describe it in mathematical terms. This is an impor-
tant part of the process of putting numbers to geology.
10 Taking the study of the Natha Jhakri tunnel to the
next stage involves a more refined numerical analysis
9
and, in order to demonstrate this process, I have used the
8
finite element program PHASE2 developed at the Uni-
versity of Toronto. This software is one of a family of
D iam eter o f plas tic z one
T u nnel diam eter

7 user-friendly but powerful programs developed with

financial assistance from the Canadian mining industry.
6 Development and distribution of these programs has
now been taken over by a spin-off company called Roc-
5 science Inc.2.
I have considered two cases, one defined by a GSI of
4
45, representing the better rock mass conditions in this
3
zone, and the other defined by a GSI of 20 that is typical
of the shear zone. As discussed earlier, the uniaxial
2 compressive strength of the intact schist is taken as
0 .0 0 .1 0 .2 0 .3 σ ci = 10 MPa and the value of the material constant mi
U niaxial c om pr ess iv e stren gth of r ock m a ss
In situ stre ss
2
Details available from Rocscience Inc., 31 Balsam Avenue,
Fig. 11: Size of plastic zone versus ratio of uni- Toronto, Ontario, Canada M4E 3B5, Fax 1 416 698 0908,
axial compressive strength of rock mass to in situ Email: software@rocscience.com, Internet:
stress. http://www.rocscience.com.
16 E. HOEK

is 10. The corresponding values of cohesion, posed on these adjacent elements causes them to fail.
angle of friction and deformation modulus, esti- The process is continued until no more elements are
mated from Figures 7, 8 and 9, are given in Table loaded to failure.
5. The uniaxial compressive strength (UCS) of the For Case 1, as shown in Figure 13, the failure zone
rock mass is calculated from the equation extends about 3 m into the rock mass surrounding the 10
UCS = 2c cos φ (1 − sin φ) and the values for the m span top heading. The convergence of the roof and
two cases are included in this table. haunches is approximately 40 mm and, in this example,
the floor heave is also approximately 40 mm. In many
Table 5: Rock mass properties for two examples cases of weak rock tunnelling, floor heave is signifi-
analysed. cantly larger than roof and wall convergence. This leads
to the need for reinforcement of the floor, by rockbolt-
Property Case 1 Case 2 ing or by the placement of a concrete invert, in order to
Intact rock strength σci MPa 10 10 stabilise the tunnel.
Material constant mi 10 10
Geological Strength Index 45 20
Cohesive strength c MPa 0.4 0.2
Friction angle φ degrees 30 23
Deformation modulus MPa 2500 550
Rock mass UCS, MPa 1.4 0.6
In situ stress MPa 6.75 6.75
UCS/in situ stress 0.21 0.09

In situ stresses along the tunnel route have been

measured by hydraulic fracturing and by over-
coring techniques (Bhasin et al 1996). The fol-
lowing values were found for the principal
stresses:
σ1 = 7.1 MPa, approximately parallel to valley,
σ2 = 5.9 MPa, vertical stress,
σ3 = 3.9 MPa, approximately normal to valley. Fig. 13: Extent of failure zone surrounding the tunnel
top heading in a rock mass defined by GSI = 45. Shear
However, because of the general weakness of failure is represented by the × symbol while tensile
the rock mass in the region of the Daj Khad shear
failure is denoted by the • symbol.
zone, it has been assumed that the rock mass can-
not tolerate significant stress differences and that
all three principal in situ stresses are equal. An
average tunnel depth of 250 m has been used to
derive the in situ stress value of 6.75 MPa used in
these analyses.
As shown in Table 5, the ratio of the uniaxial
compressive strength of the rock mass to the in
situ stress is 0.21 for Case 1 and 0.09 for Case 2.
These values fall on either side of the critical ratio
of about 0.1 shown in Figures 11 and 12.
The zone of failure for Case 1 is illustrated in
Figure 13. The PHASE2 model simulates pro-
gressive failure as the tunnel is excavated. The
process used to achieve this simulation involves
transferring loads that cannot be carried by failed Fig. 14: Reduced failure zone in the top heading roof
elements onto adjacent elements. A check is then due to the installation of untensioned fully-grouted
performed to determine whether the loads im- rockbolts and steel fibre reinforced shotcrete.
 PUTTING NUMBERS TO GEOLOGY 17

Figure 14 shows that the top heading in the drainage and grouting of the rock mass ahead of the
better quality rock mass (GSI = 45) can be stabi- tunnel.. These are followed by a 12 m long umbrella of
lised by a combination of untensioned fully- grouted pipe forepoles, forming a protective umbrella
grouted rockbolts and steel-fibre reinforced shot- under which the tunnel can be excavated. Cemented
crete. The rockbolts are 4 m long, 25 mm diame- fibreglass bars are used to stabilise the face and steel
ter and are installed on a grid pattern of 1.5 m x sets, radial rockbolts and a shotcrete or concrete invert
1.5 m. The shotcrete layer is 100 mm thick. Typi- are also used if required.
cally a 25 mm thick layer of shotcrete is placed Figure 16 shows the equipment used to drill the sub-
immediately after the excavation of a tunnel horizontal holes and to install the forepoles in the Daj
length of two to three metres. This is followed by Khad stretch of the Nathpa Jhakri headrace tunnel.
the installation of the grouted rockbolts to within The three-dimensional geometry of the tunnel head-
about 1 m of the face. A second layer of shotcrete ing and protective umbrella makes it very difficult to
is then applied to bring the total thickness up to analyse this support system. Two-dimensional analyses,
100 mm. In this case, no support of the floor is such as those described above, are not adequate. Grasso
required since this is relatively stable and it will et al (1993) used an axisymmetric two-dimensional
be excavated during the subsequent benching op- model to study the support provided by the forepole
eration. umbrella. However, I feel that a full three-dimensional
In deciding upon the adequacy of the support analysis of this support system would be justified.
system, the extent of the failure zone in the rein- Three-dimensional models capable of a full progressive
forced rock mass is checked. Rockbolts passing failure analysis for this type of support system are be-
through this failure zone will generally suffer coming available but are not for the numerically timid.
yield of the grout/steel interface. This is not a This type of analysis is best left to the numerical model
problem provided that an unyielded anchor length specialist at this stage but they should be available as
of 1 to 2 m remains outside the zone of failed general design tool within a few years.
rock, as shown in Figure 14. The deformations in
the rock mass must also be checked to determine
whether there are any sections of the excavation
perimeter that require additional support.
Note that other support systems, such as steel
sets or lattice girders embedded in shotcrete,
could also be used to stabilise this particular tun-
nel. The final choice of the support system de-
pends upon overall cost and scheduling consid-
erations.
The Daj Khad shear zone itself is character-
ised by a Geological Strength Index of approxi-
mately 20. Mining through this poor quality rock
mass results in a failure zone that extends about
15 m into the roof and floor, as illustrated in Fig-
ure 15. The size of this zone, together with the
presence of kaolin, means that rockbolt support
will not be effective in this case. Steel set support
is also difficult to design because of the large span
of the top heading and the heavy squeezing pres-
sures.
The support system chosen for mining through
this difficult stretch of tunnel is similar to that Fig. 15: Failure zone surrounding the tunnel top heading
used by Geodata on a number of previous projects in the Daj Khad shear zone, defined by a Geological
(Carrieri et al 1991, Grasso et al 1993). This con- Strength Index of 20. The tunnel convergence, shown by
sists of a series of sub-horizontal holes, up to 24 the deformed excavation boundary, is approximately
m long, for geological exploration as well as pre- 400 mm.
18 E. HOEK

Fig. 16: Installation of 12 m long grouted pipe forepoles to form a protective reinforced rock umbrella under
which excavation of the top heading can proceed.
One example of the type of three-dimensional model
that can be used for these studies is illustrated in Figure
17. This 3DEC3 model has been used in studies of the
Nathpa Jhakri underground powerhouse complex, car-
ried out by Dr B. Dasgupta of Advanced Technology
and Engineering Services, Delhi. India.

Engineering risk assessment

Fig. 17: Isometric view of the three-dimensional
numerical model of the underground powerhouse
cavern and transformer gallery of the Nathpa
Jhakri Hydroelectric Project.
The inherent variability of geological materials means
that each material property should be defined by a range
of values rather than by a single number. Hence, the end
product of any analysis based on these numbers has to
be assessed in terms of probability of occurrence or of
engineering risk.
A detailed discussion on techniques for engineering
risk assessment is beyond the scope of this paper and the
reader is referred to the excellent book by Harr (1987)
on this subject. However, the general concepts of this
3
Available from ITASCA Consulting Group Inc., Thresher
Square East, 708 South Third Street, Suite 310, Minneapolis,
Minnesota 55415, USA. Fax 1 612 371 4717
 PUTTING NUMBERS TO GEOLOGY 19

form of analysis are illustrated in the following

simple example.
The problem is to determine the risk of failure
of a slope excavated in a heavily jointed rock
mass. The shear strength properties of this rock
mass are defined by the normal distributions of
cohesion and angle of friction given in Figure 18.
These distributions were calculated by means of a
Monte Carlo simulation, using assumed normal
distributions defined by the following values
(Hoek 1998):

Parameter Mean Standard

deviation
UCS of intact rock, MPa 10 2.5
Intact rock constant mi 8 1 Fig. 19: Slope and phreatic surface geometry, rock mass
Geological Strength Index 25 2.5 properties and critical failure surface for a homogeneous
slope.
7. 0
C ohe sio n c The geometry of the slope, with a height of 60 m and a
6. 0 Me an 0 .23 MP a slope face angle of 16.7 degrees, is defined in Figure 19.
S tanda r d d ev iatio n The program SLIDE4 was used to carry out a critical
5. 0 0. 07 MP a
failure surface search, using Bishop’s circular failure
P ro ba b ility

4. 0 analysis. Rosenbleuth’s point estimate method (Hoek

1998, Harr 1987) was used to determine the mean and
3. 0
standard deviation of the normal distribution for the
2. 0 factor of the slope. This distribution is plotted in Figure
20.
1. 0

0. 0
0. 0 0. 1 0. 2 0. 3 0. 4 0. 5 4
F ac to r o f Safety
Co he s iv e s t ren gth - MP a M ean = 1.3
S tde v = 0.1
3
0 .35
F ric t ion a ngle φ
P ro ba b ility

M ea n 2 2 .9 de grees
0 .30
S tan d ard d e via tion
2
0 .25
1.3 de grees
P ro b a b ility

0 .20
1
0 .15

0 .10 0
0 .9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

0 .05

0 .00
F ac to r o f Safety
18 20 22 24 26 28
Fig. 20: Normal distribution of the factor of safety of the
F ric tion ang e - d e g rees
slope defined in Figure 19.
4
Fig. 18: Normal distributions of cohesive strength Available from Rocscience Inc., 31 Balsam Avenue, To-
and angle of friction for a heavily jointed rock ronto, Ontario, Canada M4E 3B5, Fax 1 416 698 0908,
Email: software@rocscience.com, Internet:
mass. http://www.rocscience.com.
20 E. HOEK
This plot shows that, for a mean factor of
safety of 1.3 with a standard deviation of 0.1, the
normal distribution curve extends from 0.9 to 1.7.
This range is determined by the high quality of
the input data. It was assumed that the uniaxial
compressive strength of the intact rock as well as
the material constant mi were determined by labo-
ratory testing and that the Geological Strength
Index has been obtained by careful field observa-
tions by an experienced engineering geologist.
Where poor quality input data is used for such an
analysis, the mean value may be the same but the
standard deviation and the range of factors of
safety contained in the distribution curve will be
much higher.
The probability of failure is defined by the ra-
tio of the area under the curve for factors of safety
of less than 1.0 divided by the total area under the
normal distribution curve. As can be seen from
Figure 20, this ratio is very small for the case con-
sidered. This suggests that, for this particular
slope and for the quality of the input data used, a
factor of safety of 1.3 will ensure that the risk of
slope failure is negligible.
Finite failure risks are acceptable provided
that they are considered in terms of the cost and
consequences of failure. For example, a probabil-
ity of failure of 10% may be acceptable in the
case of an open pit bench or a logging road where
traffic is restricted to trained personnel and where
equipment is available to clear up the failure. On
the other hand, this level of risk would be com-
pletely unacceptable for the abutment of a dam or
the foundation of a high rise building.
Current technology for calculating the prob-
ability of failure, as described above, can only be
used for relatively simple problems for which a
deterministic solution can be obtained. As com-
puter processing speeds increase, the application
of these methods to more complex problems, such
as the stability of underground excavations, will
become feasible.
Note that other techniques are available for
making an engineering risk assessment. These
include the use of fault and decision tree analysis
and some of these techniques are being applied to
subjects such as the assessment of dam safety
(Nielsen et al 1994). The huge societal and eco-
nomic consequences of dam failures have at-
tracted the attention of researchers in this field for
many years and we can expect to see significant
advances in risk analysis in the years to come (Anon.
1998).
Conclusion
Engineering design requires numbers. This is true
whether the design utilises man-made materials such as
steel or concrete or naturally occurring rocks and soils.
One of the principal characteristics of natural materials
is their variability and this makes it extremely difficult
to assign reliable values to the properties required by
engineering designers.
This paper has explored some of the methods that
can be used by engineering geologists and geotechnical
engineers to assess the geological factors that have an
impact on engineering design. These start from the very
crude estimates that are made during the early stages of
a project on the basis of walk-over surveys and studies
of available regional geology maps. At the other end of
the spectrum are the input requirements of the very so-
phisticated numerical analyses used to assess the stabil-
ity and support requirements for complex three-
dimensional excavations in rock.
It is easy to conclude that there is never enough in-
formation and that, what there is, is unreliable because
of the uncertainty associated with the methods of as-
signing numbers to geology. While these conclusions
may be true they are not helpful to the design engineers
who have to produce safe and economical designs,
whether or not the information is adequate.
I have tried to demonstrate that it is possible to arrive at
useable estimates of the properties required for an engi-
neering design. This requires close co-operation be-
tween engineering geologists and geotechnical engi-
neers and a good measure of common sense and practi-
cal judgement.
I would like to conclude with a statement contained
in a general report presented almost 25 years ago: ‘The
responsibility of the design engineer is not to compute
accurately but to judge soundly’ (Hoek and Londe
1974). I consider that this statement is still true today.
Acknowledgements
The permission granted by the Chuquicamata Divion of
Codelco, Chile, to publish the information contained in
Figures 1, 2 and 3 is gratefully acknowledged. Simi-
larly, permission from the Nathpa Jhakri Power Corpo-
ration, the Naptha Jhakri Joint Venture and Geodata
S.p.A. to include details of tunnelling through the Daj
 PUTTING NUMBERS TO GEOLOGY
Khad stretch of the Nathpa Jhakri headrace tunnel
is acknowledged.
I would also like to thank my wife Theo for
her support and her help in proof-reading the
manuscript of this paper.
References
Anon. 1998. Risk-based dam safety evaluations.
Workshop report. Intnl. J. Hydropower and
Dams. 5(1), 89-97, 5(2), . Denver. 1(A), 613-752.
Barton, N.R., Lien, R. and Lunde, J. 1974. Engi-
neering classification of rock masses for the
design of tunnel support. Rock Mech. 6(4),
189-239.
Bhasin, R., Barton, N., Grimstad, E. and Chrys-
santhakis. P. 1996. Engineering geological
characterisation of low strength anisotropic
rocks in the Himalayan region for assess-
ment of tunnel support. Proc. Conf. Recent
Advances in Tunnelling Technology. New
Delhi. 157-181.
Bieniawshi, Z.T. 1989. Engineering Rock Mass
Classification. Wiley, New York.
Carrieri, G., Grasso, P., Mahtab, A. and Pelizza,
S. 1991. Ten years of experience in the use
of umbrella arch for tunnelling. Proc. So-
cieta Italiana Gallerie Conf. On Soil and
Rock Improvement, Milan. March 1991,
Vol. 1, 99-111.
Fookes, P.G. 1997. Geology for Engineers: the
Geological Model, Prediction and Perform-
ance. Quarterly Journal of Engineering Ge-
ology, 30, 293-424.
Geological Survey of India. 1988. A comprehen-
sive geotechnical report of the studies car-
ried out for Nathpa Jhakri Hydel Project,
Shimla and Kinnaur Districts, Himachel
Pradesh: Part I of II and II of II. Compiled
by Ashraf, Z. and Chowdhart, A.K., New
Delhi.
Grasso, P., Mahtab, M.A., Rabajoli, G. and Pe-
lizza, S. 1993. Consideration for design of
shallow tunnels. Proc. Infrastructures
souterraines de transport, Toulon. October
1993, 19-28.
Harr, M.E. 1987. Reliability-based design in civil
engineering. New York: McGraw-Hill.
Hoek, E. 1998. Reliability of Hoek-Brown esti-
mates of rock mass properties and their im-
21
pact on design. Int. J. Rock Mech. Min. Sci. 35(1),
63-68.
Hoek, E. and Brown, E.T. 1980. Underground Excava-
tions in Rock. Institution of Mining and Metal-
lurgy, London.
Hoek, E. and Brown, E.T. 1997. Practical estimates of
rock mass strength. Int. J. Rock Mech. Min. Sci.
34(8), 1165-1186.
Hoek, E. and Londe, P. 1974. The design of rock slopes
and foundations. General Report on Theme III.
Proc. Third Congress Intnl. Soc. Rock Mech.,
Hoek, E., Marinos, P. and Benessi, M. 1998. Applica-
bility of the Geological Strength Index (GSI) clas-
sification for very weak and sheared rock masses.
The case of the ‘Athens Schist’ system of forma-
tions. Paper in preparation.
Jalote, P.M., Kumar, A. and Kumar, V. 1996. Geotech-
niques applied in the design of the machine hall
cavern, Nathpa Jhakri Hydel Project, N.W. Hima-
laya, India. J. Engng. Geol. (India) XXV(1-4),
181-192.
Nielsen, N.M., Hartford, D.N.D. and MacDonald. 1994.
Selection of tolerable risk criteria for dam safety
decision making. Proc. 1994 Canadian Dam
Safety Conference, Winnipeg, Manitoba. Vancou-
ver: BiTech Publishers, pp 355-369.
Pierson, L.A. and van Vickle, R. 1993. Rockfall Hazard
Rating System – Participants’ Manual. Federal
Highway Administration Publication No. FHWA
SA-93-057, Washington.
Russo, G., Kalamaras, G.S. and Grasso. P. 1998. A dis-
cussion on the concepts of : Geomechanical
Classes, Behaviour Categories, and Technical
Classes for an Underground Project. Gallerie e
Grandi Opre Sotterranee. March 1998, No. 54, 41-
51.
Sheorey, P.R. 1994. A theory for in situ stresses in iso-
tropic and transversely isotropic rock. Int. J. Rock
Mech. Min. Sci. & Geomech. Abstr. 31(1), 23-34.
Zoback, M.L. 1992. First- and second-order patterns of
stress in the lithosphere: the World Stress Map
Project. Journal of Geophysical Research. 97(B8),
11761-11782.